JPH04505838A - Method for determining decision results for half-image/full-image data compression methods - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] A method for determining decision results for biopsy/full image data compression methods. A method for determining the results of a full-image data compression method, which may be dynamic. For each image fragment of one image that is scanned, there is one verification value of temporally consecutive image points using the ML constant scanning method. recorded in two half-image micrographs that can be combined into a full image, and the biopsy/full image data. The data compression method is valid for the whole image when the decision result is positive, and for the two-handed image separately in other cases. Concerning how to be done.

Methods of the type described above are particularly useful for reducing data by a factor of 8 in order to reduce the required data rate. For example, when using compression, transmission is performed at a bandwidth of 120 megapints/second. It is usable for the transmission of high resolution digital television signals. At that time The data consisting of the brightness and chrominance values of the image point can be used to exploit the correlation between adjacent points. It is compressed and transmitted. For the implementation of data compression methods, for example, discrete cosacral In-conversion is applicable.

Authors Herzwinmar, Tengler, Plant, etc., published the magazine ``Nice B.I. -” No. 595, Image Coating, 1985 Title’2 Mbit/s ``New Hybrid Coding Techniques for Videoconferencing Applications'' , a television image recorded using the interlaced scanning method is coded using a discrete cosine transform. It explains the data compression method used. There, an image block of 26x26 image points is has been converted.

According to the interlaced scanning method, two half-images are recorded which are to be combined into a complete image. imperial command In the case of image fragments that do not move or are moved only slightly, the entire image is Converted for implementation. In particular, the dominant horizontal movement in the image fragment in the direction of micro-motion When moving, the images of both hands are transformed separately. At that time, the decision on fresh wound/full image compression method A method is needed to measure the results.

Authors Barbero, Strobiana, Kusi, etc. are the authors of the 2nd International Work on rHDTV. Ship J L’Aquila, 1988 paper concentration entitled rHDTV transmission Based on the DCT-based coding strategy for A method for determining the data compression method is that the data rate is It describes a method that is optimized using mean squared error.

Authors Westercumb and Beters et al. 2nd@Country on HDTV signal processing "Workshop" 1st jJ =, 5 cuola 5 superiore G, R e1s Ro+5oli, focused on papers from February 29, 1988 to March 2, 1988 with the title " Progressive scan and digital data rate reduction for future HDTV systems Conclusions for data compression methods for biopsy/full images One way to achieve this is to calculate the coefficients of non-zero in the block being transformed Explains how numbers can be used for evaluation.

Author Beters is “Image Coding Symposium PC3’8B” Turi n, Italy, 12-14 September 1988 under the title ``Component Video Biopsy/full image data pressure As a method of determining the result for the reduction method, the special coefficients of motion are estimated using luck. It describes the method used for motion vectors.

The problem of the present invention is to develop another method for determining the biopsy/full image data compression method. The purpose is to provide a method.

According to the present invention, this problem can be solved by determining the decision result as follows. resolved.

a) one equal image point in each of two directly successive micro-lines of the same half-image; For each wound-image point pair formed by two image points at the position, One half-image difference is determined from the difference in the readings of the image points of the wound-image point pair, and the difference between the two-hand images. The half-image difference for all wound-image point pairs is applied as the absolute wound sum.

b) the micro-steps that directly follow each other in the total image synthesized from the half-images from each one of the half-images; formed from two image points each at one equal image point position of the two C] Total total phase sum total image - Multiplyed as all models by the evaluation coefficient, and the gold image product is the raw wound sum The power will be reduced as a result of the decision.

The second advantageously simple method of determination is that the movement in the direction of the micro-movement can be easily determined by It is recognized by the total image sum. Based on the difference between raw wound sum and full image sum, complete image evaluation section The numbers advantageously allow movements to be evaluated in a simple manner and are also taken into account when formulating decision results. It can be done. For example, by mathematical transformation of the above relationship, the determination can also be made, for example, for fresh wounds. By a specific decision limit value for the ratio of charge removal by total image sum or raw wound sum The difference obtained by subtracting the total image sum from can be determined by the decision limit value of .

This determination method is particularly applicable to image fragments recorded using the interlaced scanning method.

The determination method is then independent of the particular data compression method. The full image evaluation coefficient is It is then applicable to data compression methods and is approximately 0.9 for most applications. Ta For example, the optimal value for the full image evaluation coefficient can be found by subjective evaluation of test tone. can be done.

Advantageously, the present determination method is not tied to any particular data compression method. day Data compression methods are, for example, discrete cosine transform, or vector quantization, or so-called A master image is used and then the image fragments are displayed in a weighted superposition. or hierarchical methods, e.g. blocking by different block sizes. This is a quadtree method in which the block is divided into internal blocks in a hierarchical manner.

In a preferred embodiment of the invention, the data compression method is a discrete cosine transform, and the It is characterized by being advantageously applicable to image contents that occur often and frequently. .

Another preferred embodiment of the invention provides that the full image evaluation factor is between 0.8 and 1.0. It is characterized by having the following. Compressed and recorded images and uncompressed images Determine the so-called signal-to-noise ratio for a large number of test tones when comparing image points with The mean squared error determined by the equation is highly desirable.

Another preferred implementation of the invention is that the full image evaluation coefficient has a value of 0.9. It is characterized by Test tone, for example "Kiel Port J" or "Puppet's Subjective" It is felt that this full-image evaluation coefficient is optimal as a basis for evaluation. Said test tone was taken by the Radio Technology Research Institute. Also, they are often less desirable than subjective evaluations. considered to be of good character.

Another preferred embodiment of the invention provides that the image fragment is a 16X16 image spot block of a television image. lock and it can be converted as one 16X16 full-image pro-nk or or separately as two 16×8 half-image blocks. especially discrete This block division is advantageous for the general cosine transform.

Another preferred embodiment of the invention provides that the image fragments are produced using high-resolution television techniques, so-called Particularly high image quality, characterized by being recorded using “high-precision television” The known advantages of this television method with respect to

Another preferred embodiment of the invention provides that the image fragment is an 8x8 image point brochure of a television image. block, which can be converted as one 8x8 full-image block or separately A particularly isolated This block partitioning is advantageous for the dispersive cosine transform. especially a statue with a lot of detail This is advantageous for the scene.

A particular advantage of the determination method according to the invention is, inter alia, that this determination method It's unrelated to. For example, with increasing contrast, the total image sum and the same ( As the sum of raw wounds increases, the decision outcome remains the same.

The invention will be explained in more detail by means of the drawing, which shows one embodiment.

FIG. 1 shows the arrangement for this embodiment.

FIG. 2 shows an image fragment of this embodiment.

FIG. 3 shows the marked readings of the first half-image.

FIG. 4 shows the marked readings of the second half-image.

Figure 5 shows the readings filled in for the entire image.

Figure 6 shows the numerical values entered for the readings of the whole image.

FIG. 7 shows a flow diagram for image fragment compression.

As shown in FIG. 1, in this embodiment with camera 3000, the horizontal An image fragment 1000 moving to the left is recorded. At that time, the inspection recorded using the interlaced scanning method The values are compressed and stored in the summer measurement device 4000 after performing a discrete cosine transform. It will be done.

As shown in FIG. 2, the image fragment 1000 consists of 16×16 image points. Ru. The image fragment 1000 is recorded in two half-images using an interlaced scanning method. first recorded The hand wound consists of micro lines 101, 102 to 108. The second move that follows The wound consists of minor lines 201, 202 to 208, 1 amphipathic wound, minor line 301 ．． The entire image consisting of 302 to 308 is combined. The object to be imaged is that The diagonal border between the bright region 900 and the dark region 800 of the image fragment 1000 It is. This boundary line moves horizontally to the left, and as it moves, image fragments are recorded】0 00, a boundary line 600 is obtained.

Without this movement, a boundary line 500 would be obtained in the recorded image fragment 1000. This boundary line is drawn in dashed lines for ease of understanding.

The reading value of one image point within the bright range 900 is in this example determined by a value equal to 1. will be recorded. The reading value of one image point in the dark range 800 is O in this example. Recorded with equal valence. The reading value Bj+1 recorded at the image point is distinguished by the subscript. It will be done. The first subscript j indicates the image point position in micro-rows 301 to 316, respectively. The subscript 2 indicates the position of the slightest line in the entire image that can be synthesized from the hand wound. In this way For example, in this embodiment, the reading recorded at the 15th image point position in the 16th image row B1%++i-0 and recorded at the 16th image point position in the 16th image row The measured value B is 11+16=1.

As shown in FIG. 3, the reading of the first hand wound in this example is The data structure 2901 is divided into 16 rows and 16 columns like 1000. Crab is written. The raw wound-image point pair P, l corresponds to each of the two points of directly successive micro-motions. Each image point is formed from two image points at one equal image point position.

As shown in FIG. 4, the reading of the second hand wound in this example is 1000, the data structure 2902 is divided into 16 rows and 16 columns. Crab is written. The raw wound-image point pair P is in each of the two directly succeeding micro-motions. are formed from two image points in each case at one equal image point position.

As shown in FIG. 5, the image value of the entire image in this example is 100 It is written in data structure 2903, which is divided into 16 rows and 16 columns like 0. be included. The whole image ~ Image point pair Pv is each one of two directly successive micro-steps. are formed from two image points at equal image point positions, respectively.

As shown in FIG. 6, in the data structure 2903 in this embodiment, The numbers O and 1 are written as readings.

The raw wound sum HS is calculated in this example as follows.

The total phase detection ■S is calculated as follows in this embodiment.

Then, when the decision result E uses a full image evaluation coefficient of 0.9, it is calculated as follows. It will be done.

E-MS-0,9・VS--0,1 According to the negative decision result, the data compression method is applied separately to both wounds in this example. Ru.

Therefore, in this embodiment, due to the difference in the raw wound sum and the total phase inspection, the horizontal direction of the micro-travel direction is The motion recognized by the full-image evaluation coefficient and the In this example, the correlation between adjacent points in the whole image is It is done so that the correlation of adjacent points is given priority during data compression. achieved.

As shown in FIG. 7, in this embodiment, in the computing device 4000, the main The calculation program shown in the flow diagram illustrates the progress according to system is used.

Starting at program point 2000, the reading of the first hand wound is first decoded by command 2001. data structure 2901.

Thereafter, instruction 2002 writes the reading of the second hand wound into data structure 2902. be included.

After that, the reading of the entire image is written into the data structure 2903 by instruction 2003. Ru.

Thereafter, the raw wound sum is calculated by instruction 2004. At that time, a fresh wound from both sides - image The pruritus image differences formed from the differences in the readings of the pair of points are added absolutely. In this example, A value of 26 is calculated for 1000 image fragments for the raw wound sum.

Thereafter, instruction 2004 calculates the entire phase detection. At that time, from the whole image to the whole image - image point The total image differences formed from the differences between the paired readings are added absolutely. In this example, The value 29 is then calculated for the image segment +000 for the entire phase detection.

Instructions 2006 then calculate the decision result. At that time, all inspections will be carried out in this manner. In the example, it is multiplied by a full image evaluation coefficient equal to the numerical value 0.9. formed at that time All models are subtracted from the raw wound sum, so in this example for 1000 image fragments 10, the decision result having a numerical value equal to 1 is selected.

The decision result is then checked by instruction 2007, and in case of a positive decision result, A branch is made to instruction 2009 which performs a discrete cosine transform on the entire image.

In this example, in the event of a negative decision result, as is the case for image fragment 1000, Instructions 2008 perform a discrete cosine transform on both wounds separately.

In both cases the calculation program then continues at program point 2010. Ru.

FIG3 FIG4 IG5 IG 6 Figure 7 international search report SA　35333

Claims (7)

[Claims] 1. A method for determining a decision result for a half-image/full-image data compression method, the method comprising: From the image fragments of one image, which are moved if necessary, successive images in time in an interlaced scanning manner The image values of a point are recorded in the image rows of two half-images that can be combined into one full image; The data compression method for image/full image is applied to the whole image in case of a positive decision result, and separately in other cases. In the method applied to both hemi-images, the decision result is as follows: a) the same One of each of the two immediately coarse following image rows (100, 200) of the half-image of one Each half image formed from two image points at equal image position - image point One half-image difference for each pair (PM) is the image value of the image point of the half-image-image point pair (PM) (BjM) and all half-image-image point pairs (PM) of both half-images. b) From each one of the half-images, the half-image differences are summed absolutely as a half-image sum (HS) 2 of each of the directly successive image rows (300) in the full image synthesized from the half-images of each one formed from two of the image points at one equal image point position in the For each full image-image point pair (PM), one total image difference is obtained for each full image-image point pair (PV ), and all image-image point pairs (PV ), the total image difference is absolutely added as the total image sum (VS), c) The full image sum (VS) is multiplied by the full image evaluation coefficient as the full image product, and the full image product is A half-image/ A method for determining decision results for full-image data compression methods. 2. Claim 1 characterized in that the data compression method is discrete cosine transformation. How to put it on. 3. Claim characterized in that the full image evaluation coefficient has a numerical value between 0.8 and 1.0. The method according to scope 1 or 2. 4. The person according to claim 3, characterized in that the full image evaluation coefficient has a numerical value of 0.9. Law. 5. An image fragment is a 16x16 image point block of one television image, which is one converted as two 16x16 full image blocks or two each separately 16 x 8 half-image blocks of 4. The method described in one of 4. 6. The image fragments are produced using high-resolution television methods, so-called “high-precision television.” 6. The method according to claim 1, wherein the method is recorded. 7. An image fragment is an 8x8 image point block of one television image, which is one converted as 8x8 full image blocks or separately into two 8x4 According to one of claims 1 to 6, characterized in that it is transformed as a half-image block. Method described.